A carve-out gets underwritten like any other deal. Price, structure, synergies, a return that works if the model holds. The diligence is thorough on all of it. Then the deal closes, and the thing that actually determines whether the model holds turns out to be a question the model barely touched: can NewCo stand up the functions the parent used to provide, before the clock the seller set runs out.
That's not a financial question. It's an operational one, and underneath it, a talent one. It's also the part of a carve-out that gets budgeted last, if at all.
The one deal where the functions don't exist yet
Every other kind of buyout hands you a running company. The finance team, the IT function, the HR group, they come with the business. Calibrated to the old owner, maybe wrong for where you're taking it, but present and running on day one. Whatever else you're fixing, the lights stay on while you fix it.
A carve-out doesn't work that way. The business you bought ran on the parent's functions. Finance closed the books on the parent's systems, run by the parent's people. IT lived in the parent's environment. Procurement, HR, treasury, reporting, all of it sat at the center and served the unit you're now carving out. When the deal closes, almost none of that comes with it. The systems stay with the parent. The people who ran them stay with RemainCo, because that's their employer and their career. NewCo gets the P&L, the assets, and a transition services agreement. It does not get the operators.
So you don't inherit functions that need fixing. You inherit an obligation to build functions that aren't there, while running the business at the same time.
The TSA is a countdown, not a cushion
The transition services agreement feels like protection, and in the narrow sense it is. The seller keeps the lights on for a defined window while you get established. But the TSA is a deadline wearing the costume of a safety net, and treating it as the second thing is how carve-outs slip.
The seller has no incentive to make the arrangement comfortable or permanent. Rates are set to encourage your exit, extensions are expensive by design, and the parent wants its shared services back to serving itself. Every function sitting under that TSA has to be rebuilt inside NewCo and cut over before the window closes. Systems migrated. Processes stood up. A team in place to run each one. Miss the date and your options are both bad: pay a premium to extend, or operate past the deadline on infrastructure you no longer have clean rights to. Value leaks either way, and it leaks on a schedule you agreed to at signing.
The model prices the business and underbudgets the build
Here's where it goes wrong on paper before it goes wrong in practice. The deal model prices the business to the dollar. Synergies get their own tab. The one-time cost to actually separate, and specifically the talent to execute that separation, rarely gets modeled with the same rigor. Separation shows up as a line for advisory fees and systems spend, and the people who have to own the work are assumed into existence.
Then the calendar does what calendars do. Month three arrives and no one owns the systems migration end to end. The first standalone close is two months out and there's no controller who has ever run one for this entity. The reporting the board expects doesn't have an owner, because the person who would own it is a hire that hasn't started because the search hasn't finished. The scramble begins with the TSA clock already down a third, which is the worst possible time to start, because now every solution is the expensive one.
Build against the clock the TSA set
The carve-outs that hold the model treat separation as a build with a hard end date, and they resource it as one from Day 1 rather than discovering the need in month three.
Where a function that lived at the parent now has to be owned at NewCo, an interim leader owns it from the first day. An interim finance chief runs the standalone finance function while you work out what the permanent one looks like, which you can't know at close because the standalone entity doesn't exist yet in any real form. The seat is covered and accountable while the permanent picture develops, instead of sitting open while a search runs against the TSA clock.
The separation workstreams themselves are finite and specialized, which is exactly the on-demand case. Systems cutover. The TSA exit plan and the sequencing of which services come off when. Standing up reporting from nothing. Getting to a first standalone close that ties out. These are projects with a hard deadline and a defined end, run by specialists who have done the specific thing before, for as long as the separation runs and no longer. They don't become headcount on a company still learning its own shape.
The permanent hires come when that shape is finally visible. Once the functions are standing and the standalone entity is operating as itself, you know the team it actually needs, and you hire against a real picture instead of a guess made in a data room. Try to make those permanent hires at close and you're hiring for a company that doesn't exist yet, which is how you end up with the wrong team for the entity that emerges.
Day 1 readiness is a talent plan
The instinct is to treat separation readiness as a systems problem, and the systems are real. But a system with no one to own it is a liability, not an asset, and the difference between a carve-out that hits its model and one that bleeds through TSA extensions is usually not the technology. It's whether someone was accountable for each function from Day 1, and whether the finite separation work had specialists pointed at it while the permanent org came into focus.
That's a talent plan, and it belongs in the deal model, priced and sequenced before close, not assembled under pressure once the clock is already running. The separation is going to cost what it costs. The only choice is whether you budget it in the data room or pay the premium for it in month four.